Ok, we have this figured out.
We have a cluster, 10.17.6.167
In that cluster we have a port, 80
In port 80 we have four "Server" documents: GDDATA, GDDATA2, GDDATA3,
GDDATA4.
All four of these servers have the same "HTTP Advisor Request": GET
/gdintranet.nsf/edge HTTP/1.0
Each one of these servers has an "HTTP Advisor Response" crafted to it:
GDDATA uses GDDATA, GDDATA2 uses GDDATA2 and so on.
If I bypass the cluster and go to
http://gddata/gdintranet.nsf.edge I will
see only GDDATA, GDDATA2 and I will see only GDDATA2 and so on.
Port 80 also has four "Rule" documents. These are "Always true" rules.
GDDATA has priority 1, GDDATA2 has 100, GDDATA3 has 50. Actually there is
no GDDATA4 rule as it's server doc is flagged down.
The "Manager" has an Advisor document, Advisor: Http 10.17.6.167:80. It
has "Minimal" logging in use. Verbose helped us with that "HTTP Advisor
Request".
If I tell http to quit on GDDATA it knows and starts going to the other
servers. If I inject a typo in the "HTTP Advisor Respose" (like change
GDDATA to GDxyzDATA) then it will use the other systems instead also.
So now when my users go to
http://dekkohome.corp.dekko.com/... they will
get routed to one of the GDDATA servers based on whether or not accessing
port 80 returns the right system name.
Our dns points dekkohome.corp.dekko.com to 10.17.6.167.
Things are good.
Rob Berendt
midrange.com
